Division of Neurology, Barrow Neurological Institute, St. Joseph "s Hospital and Medical Center, Phoenix, Arizona

Transcription

1 J Neurosurg 76: l, 1992 A randomized comparison of intra-arterial versus intravenous BCNU, with or without intravenous 5-fluorouracil, for newly diagnosed patients with malignant glioma WILLIAM R. SHAPIRO, M.D., SYLVAN B. GREEN, M.D., PETER C. BURGER, M.D., ROBERT G. SELKER, M.D., JOHN C. VANGILDER, M.D., JAMES T. ROBERTSON, M.D., JOHN MEALEY, JR., M.D., JOSEPH RANSOHOFF, M.D., AND M. STEPHEN MAHALEY, JR., M.D., PH.D. Division of Neurology, Barrow Neurological Institute, St. Joseph "s Hospital and Medical Center, Phoenix, Arizona u- This Phase III trial tested the efficacy and safety of intra-arterial 1,3-bis(2-chloroethyl)-l-nitrosourea (BCNU) for the treatment of newly resected malignant glioma, comparing intra-arterial BCNU and intravenous BCNU (200 mg/sq m every 8 weeks), each regimen without or with intravenous 5-fluorouracil (l gm/sq m three times daily given 2 weeks after BCNU). All patients also received radiation therapy. A total of 505 patients were randomly assigned within the study. Fifty-seven patients were excluded, primarily because of neuropathology error, and the remaining 448 patients constituted the Valid Study Group. Of the total 505 patients, 190 patients could not receive intra-arterial BCNU and 315 patients were randomly assigned to receive intra-arterial (167 patients) and intravenous (148 patients) BCNU. Actuarial analysis (log-rank) demonstrated reduced survival for the intra-arterial group (p = 0.03). Serious toxicity was observed in the intra-arterial group; 16 patients (9.5%) developed irreversible encephalopathy with computerized tomography evidence of cerebral edema, and 26 patients (15.5%) developed visual loss ipsilateral to the infused carotid artery. Administration of 5-fluorouracil did not influence survival. The survival rate between the intravenous and the intra-arterial BCNU patients with glioblastoma multiforme did not differ, but was worse for intraarterial BCNU patients with anaplastic astrocytoma than for those receiving intravenous BCNU (p = 0.002). Neuropathologically, intra-arterial BCNU produced white matter necrosis. It is concluded that intra-arterial BCNU is neither safe nor effective in prolonging survival when administered by the methods used in this study of newly diagnosed patients with malignant glioma. KEY WORDS brain neoplasm chemotherapy multicenter study 9 intra-arterial BCNU radiation therapy I N 1983, the Brain Tumor Cooperative Group (BTCG), previously named the Brain Tumor Study Group, reported that surgery plus radiation therapy and chemotherapy with intravenous 1,3-bis(2-chloroethyl)-l-nitrosourea (BCNU, carmustine) significantly added to the survival time of patients with malignant glioma in comparison to those undergoing surgery plus radiation therapy without chemotherapy. 2~ Patients with all three modalities had a median survival period of 1 year, whereas those with surgery plus irradiation only survived a median of 10 months. The 18-month survivorship was 2.5-fold greater among the patients receiving carmustine than among those without chemotherapy. This and subsequent studies confirmed the value of intravenous BCNU, and showed that neither procarbazine nor streptozotocin was more effective. ~5 Despite these results and subsequent additional experience with intravenous BCNU, 53 the drug was clearly not curative and methods were sought to improve its efficacy. Evidence from both the laboratory 4~ and the clinic 23"z4" :6,27.42 suggested that higher concentrations of BCNU may be achieved in tumor by regional chemotherapy consisting of intra-arterial drug infusion. Several trials in small numbers of patients reported clinical efficacy of intra-arterial BCNU, and several made note of what was believed to be an acceptable low incidence of eye toxicity and encephalopathy. 6'lg'2~176 eye toxicity consisted primarily of pain in association with the intra-arterial infusions delivered into the in- 772 J. Neurosurg. / Volume 76 / May, 1992

2 Intra-arterial BCNU for malignant glioma ternal carotid artery (ICA) below the takeoff of the ophthalmic artery. Visual loss was occasionally reported, thought to be a result of direct retinal toxicity from the drug. The encephalopathy occurred over a few days and was characterized by progressive dysfunction in the part of the brain that was supplied by the infused artery. Pathological examination demonstrated necrosis in the white matter that had many of the hallmarks of radiation damage. However, the incidence of encephalopathy was low and was usually associated with doses of intra-arterial BCNU exceeding 400 mg/sq m. 24 To test the assumption of efficacy and safety, in December, 1983, the BTCG began accrual of data to Trial 8301, a randomized Phase III trial comparing intra-arterial BCNU versus intravenous BCNU, each regimen with or without intravenous 5-fluorouracil. As noted below, intra-arterial BCNU was found to be too toxic to the brain, and patient assignment to the BCNU intra-arterial group was stopped in October, 1986; accrual to the remaining arms of the study ended in March, Because of the finding of brain toxicity, preliminary reports were published warning of the danger of the regimen. 44'52 We now present a final analysis of the data from the trial. Patient Accrual Clinical Material and Methods Seven institutions participated in the design and conduct of this trial (see Appendix). Patients were entered in the study by the principal investigators at each institution within 3 weeks after definitive surgical treatment for the primary tumor. Eligibility criteria included: the presence of a histologically demonstrated supratentoriai malignant glioma; a patient age of 15 years or greater; the absence of other major illness that might preclude treatment on any arm of the study; and a Karnofsky Performance Status score of 40 or greater at the time of assignment. Eligible patients were not to have received any previous antineoplastic therapy, but conventional doses of corticosteroids were allowed. Written informed consent was obtained from all patients. Before assignment to a chemotherapy group, the best possible conventional neurosurgical procedure was performed. In particuiar, maximum tumor resection and internal cerebral decompression commensurate with good medical judgment were accomplished in all patients. A set of pathology slides was sent to the BTCG Operations Office for subsequent evaluation by the Neuropathology Review Committee, which made the final determination of eligibility on the basis of defined histopathological criteria. 5 Radiation Treatment Radiotherapy was begun within 3 weeks after surgical resection in conjunction with chemotherapy as prescribed below. Initial radiation was delivered to the whole brain and consisted of 4300 cgy in 25 fractions of 172 cgy each over 5 weeks via bilateral opposing ports. The whole-brain portion of the radiation protocol was followed by a coned-down boost of an additional 1720 cgy, for a total tumor dose of 6020 cgy (approximately 1700 rets), delivered to the tumor with the 90% isodose contour encompassing the target volume plus a 2-cm margin. The coned-down volume was determined by defining the volume of the tumor on computerized tomography (CT) scans prior to initiation of radiation therapy and allowing for a margin of at least 2 cm, but sparing as much brain as possible. Chemotherapy The study had a factorial design. Patients were randomly assigned to receive either intravenous or intraarterial BCNU (200 mg/sq m) once every 8 weeks. Patients in each of these two groups were also randomly assigned to receive BCNU alone or BCNU followed 14 days later by the addition of intravenous 5-fluorouracil (1 gm/sq m/day) for 3 days, every 8 weeks. Such patients, randomly assigned with equal probability to one of the four groups in this 2 x 2 factorial design, were considered part of the "full randomization." Some patients were deemed on medical grounds not to be eligible to receive the intra-arterial BCNU, usually because of ultrasound or arteriographic evidence of severe arteriosclerosis in the artery to be infused. These patients were to receive intravenous BCNU and, in addition, were randomly assigned to receive or not receive 5-fluorouracil. These patients, eligible only for one of the two groups receiving intravenous BCNU, were considered part of the "limited randomization," and contributed to the evaluation of 5-fluorouracil. Also included in the limited randomization were all patients assigned after October, 1986, when accrual to the intra-arterial BCNU group was halted for reasons described below. The initial dose of chemotherapy was to be delivered within 72 hours after beginning radiation therapy. Intravenous BCNU was administered in a single dose of 200 mg/sq m over 30 to 60 minutes. According to protocol, this dose of BCNU was repeated at 8-week intervals if the platelet count nadir was not below 50,000/ cumm, the white blood cell (WBC) count nadir was not below 2000/cu mm, or the hematocrit drop was not greater than 10 points, and if the recovery platelet count was over 100,000/cu ram, the WBC count was greater then 3000/cu ram, and the hematocrit was greater than 30%. A reduced BCNU dose of 150 mg/sq m was administered if the nadir counts reached the limits described above, and recovery exceeded a platelet count of 75,000/cu mm, a WBC count of 2500/cu ram, and a hematocrit of 28%. A further reduction to 100 mg/sq m was ordered if the same indicators of hematotoxicity were observed after treatment at a reduced dose. If the nadir counts mentioned above were reached after delivery of BCNU doses of 100 mg/sq m, no further BCNU was administered. Intra-arterial BCNU was administered on the above schedule through a catheter placed into the femoral artery under local anesthesia in the neuroradiology suite. The arterial supply of the tumor was confirmed prior to J. Neurosurg. / Volume 76/May,

3 W. R. Shapiro, el al. BCNU infusion, and the tip of the catheter was placed in the appropriate artery. If the tumor was located in the posterior portion of the cerebral hemisphere and was determined to be supplied wholly or in part by the vertebral artery, the BCNU dosage was split between the ICA (two-thirds of the dose) and the vertebral artery (one-third of the dose). Tumors of the corpus callosum were infused as follows. If the bulk of the tumor was located in one hemisphere, the ICA supplying that hemisphere was cannulated and the total dose was infused. If the tumor was distributed approximately equally between the two hemispheres, a full dose of BCNU was given via one ICA at one administration and the other ICA at the next. Infusions of subsequent doses were alternated between the two sides. For ICA infusion, the tip was located above the common carotid artery bifurcation; no attempt was made to pass the tip to a point above the ophthalmic artery takeoff point. For vertebral infusion, the tip was located in the appropriate vertebral artery. The BCNU was made up in the standard fashion, dissolving I00 mg of the drug in 3 ml of absolute ethanol, to which sterile water for injection was added to make the total 30 ml. As noted below, burning eye pain frequently accompanied ICA infusions of BCNU, and consideration was given to deleting the ethanol and using only water for injection with warming and vigorous shaking, as has been reported. 29'36 However, drug recovery experiments using the method of Krull, et al., 3~ demonstrated that as much as 35% of the BCNU failed to dissolve in the water without the ethanol. Furthermore, the deletion of ethanol was not reported to lower the incidence of eye pain appreciably, and it was elected to continue the standard dilution schedule with ethanol. Although the BCNU was scheduled to be infused at 4 ml/min, if the eye on the ipsilateral side became painful during ICA infusions, the infusion was slowed to 2 ml/min. Patients were also given narcotic analgesics, and ice was applied to the eye. Overall, these measures usually permitted the infusions to be completed with only mild to moderate distress to the patient. During the course of the study, a serious toxicity occurred in the intra-arterial group. By May, 1985, 13 patients had developed encephalopathy. All 13 patients had received at least two courses ofintra-arterial BCNU, and the encephalopathy was documented in the 3rd to 7th month of therapy. Because of the occurrence of encephalopathy, the BTCG acted in May, 1985, to decrease the dose of intra-arterial BCNU to 100 mg/sq m beginning in Course 3, with the remaining 100 mg/ sq m given by the intravenous route. In July, 1985, it was decided to change Course 2 likewise. Because of the potential for pulmonary fibrosis with large cumulative doses of BCNU,~ pulmonary function studies were performed prior to the initial dose of BCNU and were repeated at regular intervals as described. 53 In addition, no further BCNU was given after the course of treatment that brought the total cumulative amount of drug in excess of 1500 mg/sq m. The carbon monoxide (CO) single-breath diffusion test was routinely utilized. If at any time the results of subsequent CO diffusion tests indicated a decrease of 15% or more from the baseline value (% of predicted), then treatment with BCNU was discontinued. A course of 5-fluorouraeil was administered 14 days after the BCNU. This was given at a dose of 1 gm/sq m/ day by continuous intravenous infusion over 3 days. No alteration of 5-fluorouracil dosage was required based on the hematological response; only the dosage of BCNU was modified. Ancillary Treatment Anticonvulsant medications were used as necessary to control seizures. Corticosteroids were used in all treatment groups as needed for the control of cerebral edema. Statistical Analysis Comparisons of periods of survival were made among the several groups. The probability of survival was calculated for a specific length of time by the lifetable method ~ and compared with the survival rate of different groups using a Mantel (log-rank) statistic, 45 always with two-tailed tests. In addition, survival times of the treatment groups adjusted for important prognostic factors were compared using a Cox proportionalhazards model? ~ Testing for significance of treatment effects was performed using the coefficient for treatment divided by its standard error. Score statistics were used to identify possible important interaction terms, which were tested formally as above. During the course of the trial, interim data monitoring showed a trend toward better survival for the intravenous BCNU group compared to the intra-arterial BCNU group, although at the time the difference was not statistically significant. In view of the intra-arterial BCNU toxicity that had been recognized, an independent review group sponsored by the National Cancer Institute was convened and an early stopping strategy was tested. Approximate calculations were performed to indicate whether this trial might eventually show a statistically significant result in favor of intra-arterial BCNU. It was determined that detection of a 50% increased survival period for the intra-arterial BCNU group would have had only a 2% chance even with 300 deaths. For this reason, the intra-arterial BCNU trial was stopped. Patient Population Results A total of 505 patients were involved in the study, ranging in age from 18 to 79 years (median 56 years). At the time of analysis, 409 of the 505 patients had died; the median follow-up time for the surviving 96 patients at last contact was 43 months (range 0 to 83 months). As is appropriate for a randomized trial, we analyzed the data for the total population of 505 patients to avoid the bias that is possible when patients 774 J. Neurosurg. / Volume 76/May, 1992

4 Intra-arterial BCNU for malignant glioma TABLE 1 Characteristics ~?[patients in the Valid A'ludy Gr~tq~* Full Randomization Group Full + Limited Characteristic Groups IV BCNU IA BCNU Total No 5-FU 5-FU Total no. of cases age at randomization (%) yrs yrs yrs yrs KPS score (%) histopathology (%) GBM non-gbmt * IV = intravenous; 1A = intra-anerial; 5-FU = 5-fluorouracil" KPS = Karnofsky Performance Status; GBM = glioblastoma multiforme. "f The non-gbm category of the Valid Study Group consisted of 102 cases of anaplastic astrocytoma, six of anaplastic oligodendroglioma, and four of malignant mixed glioma. are excluded from analysis. We also analyzed separately those 448 patients who were subsequently confirmed to have met protocol eligibility specifications (including the central pathology review), termed the "Valid Study Group." The reasons for exclusion were as follows: the neuropathology review of slides revealed that 40 patients did not have malignant glioma; 15 patients did not have adequate material available for neuropathology review; and two patients were ineligible for the study (one because radiotherapy was received more than 1 week before randomization and the other because of a low platelet count). Three of the patients without neuropathology review did subsequently have a successful review and could therefore have been included in the Valid Study Group, but this would have had a negligible effect on the major conclusions. The characteristics of the Valid Study Group are listed in Table 1 by randomized treatment group. It is of note that 75% of these patients had glioblastoma multiforme. For comparing intra-arterial BCNU with intravenous BCNU administration, the analysis was restricted to the 315 patients eligible for the full randomization (the full factorial design), because the comparison of the two BCNU regimens is properly done only for those patients for whom the decision was made by random assignment. The remaining 190 patients in the limited randomization (eligible for only one of the BCNU intravenous groups) may not be comparable to the others and could theoretically bias the comparison between intra-arterial and intravenous BCNU administration. However, all patients were randomly assigned to receive or not to receive 5-fluorouracil, so all patients were included in the evaluation of 5-fluorouracil. FIG. 1. Graph showing patient survival time from randomization correlated with the chemotherapy group for the 279 patients eligible to receive intra-arterial (IA) BCNU in the Valid Study Group. Numbers in parentheses denote numbers of patients in each group. IV = intravenous; 5-FU = 5-fluorouracil. Survival Rate for the Valid Study Group We first considered survival time (from time of assignment) by treatment group limited to patients in the full randomization (Fig. 1). A test for overall heterogeneity across the four curves yielded p = It is more relevant to consider the same data in two categories, grouped by random assignment of intra-arterial versus intravenous BCNU administration. The difference was statistically significant (p = 0.03), with the intra-arterial BCNU group having a shorter survival. As seen in Fig. 2 upper, the median survival times were 11.2 months for the intra-arterial BCNU group and 14.0 months for the intravenous BCNU group, with 13% and 25% probabilities of surviving 2 years, respectively (life-table estimate). Knowing that difficulties occurred more often in administering intra-arterial than intravenous BCNU, we also performed a subsidiary analysis, excluding the 19 patients who did not receive at least one course of the BCNU regimen to which they were assigned. The result actually showed a slightly greater difference than before, again with the intravenous BCNU group having a longer survival (p = 0.008). All patients, whether in the full or limited randomization, contributed to the analysis of survival data in the assigned 5-fluorouracil group. The differences were not statistically significant (p = 0.96), and the curves were practically superimposed, as seen in Fig. 2 lower. Prognostic Variables in the Valid Study Group Analyses of other variables in these data show (Fig. 3) that the histopathological category, age at randomization, and Karnofsky Performance Status score at randomization were all markedly significant prognostic variables (p < for each variable considered separately). We applied a Cox proportional-hazards model to study survival data in the Valid Study Group, incorporating these three prognostic variables as well as indicator variables for the BCNU assignment (intravenous J. Neurosurg. / Volume 76 /May', I

5 W. R. Shapiro, et al. FIG. 2. Graphs showing survival time for the Valid Study Group from randomization correlated with chemotherapy.groups. Numbers in parentheses denote numbers of patients m each group. Upper: Correlation with randomized assignment of intra-arterial (IA) versus intravenous (IV) BCNU for patients eligible for the full randomization (half of each group also received 5-fluorouracil). Lower." Correlation with patient's randomized assignments to receive 5-fluorouracil (5- FU) or no 5-FU treatment. vs. intra-arterial), for the 5-fluorouracil assignment (no vs. yes), and for the randomization option (full vs. limited). As with the previous BTCG study (Trial 8001), the results indicated that, although age was an important prognostic factor in both histopathological subgroups, the effect of age was more pronounced in the nonglioblastoma patients, so an interaction term was included to model this effect. Significance testing was based on a coefficient divided by its standard error. The three prognostic variables were all significantly predictive when included together in the model, but the treatment variables were not statistically significant. For 5-fluorouracil, this was consistent with the unadjusted analysis. For BCNU, the model did indicate a worse survival for the intra-arterial group, with an estimated relative-hazard rate of 1.2, but this was not statistically significant (p = 0.18). Interestingly, the model did indicate a statistically significant (p = 0.003) interaction between the BCNU assignment group and histopathology. Figure 4 illustrates this interaction, showing similar survival for the two BCNU groups (intra-arterial slightly shorter) among the glioblastomas, while the intra-arterial group had considerably shorter survival among the nonglioblastomas. Indeed, this shortened survival time among the intra-arteriai BCNU-treated nonglioblastoma patients contributed most of the difference FIG. 3. Graphs showing survival time for the Valid Study Group from randomization correlated with the histopatbological diagnosis (upper), patient age at randomization (center), and Karnofsky Performance Status score at randomization (.lower). Numbers in parentheses denote numbers of patients in each group. to the overall worse survival time of patients treated with intra-arterial BCNU (Figs. 1 and 2 upper). Survival Time for the Randomized Population To avoid any possibility of bias produced by the exclusion of patients, we analyzed survival data by treatment group for the total population of 505 individuals. The results were qualitatively similar to those for the Valid Study Group. Toxicity Adverse events representing possible toxicity were summarized over all courses of chemotherapy received by the study patients and are presented separately by treatment group assignment (Table 2). In addition to 776 J. Neurosurg. / Volume 76/May, 1992

6 Intra-arterial BCNU for malignant glioma TABLE 2 Summao, of adverse evenl~ encountered in each treatment group* Treatment Group F{G. 4. Graphs showing survival time for the Valid Study Group from randomization correlated with chemotherapy groups (intra-arterial (IA) vs. intravenous (IV) BCNU) for patients eligible for IA BCNU. Numbers in parentheses denote numbers of patients in each group for glioblastoma multiforme cases (upper) and nonglioblastoma (mostly anaplastic astrocytoma) cases (lower). specific data on clinical hematology and chemistry, qualitative information was collected on the incidence of a variety of adverse medical sequelae during chemotherapy according to a checklist submitted to the Operations Office; this information, although not specifically documented, is useful as a general indication of toxicity when comparing the three groups. When comparing systemic toxicity of the intra-arterial versus intravenous BCNU administration, it should be noted that on average the intra-arterial groups received fewer total courses of chemotherapy. Table 2 also shows the adverse reactions attributed to intra-arterial chemotherapy, including 16 patients experiencing encephalopathy. The incidence of adverse reactions appeared similar both with and without the use of 5-fluorouracil. As noted above, the most serious adverse reaction associated with intra-arterial BCNU was encephalopathy. The symptoms developed only after two or more courses of intra-arterial BCNU, and were associated with or preceded by severe ocular toxicity progressing to visual loss in a majority of patients (11 patients had both encephalopathy and visual loss, five had encephalopathy alone, and 15 others had visual loss). The encephalopathy was manifested by the subacute development of repeated seizures, hemiparesis contralateral to the perfused hemisphere, and progressively deteriorating mental status. The clinical picture resembled Adverse Event IV IV BCNU IA IA BCNU BCNU + 5-FU BCNU + 5-FU (%) (%) (%) (%) WBC < 2000/cu mm platelets < 50,000/cu mm , [ hematocrit < 25% 2, abnormal liver functioni abnormal renal functiont , pulmonary fibrosis infection severe nausea/vomiting/ diarrhea related to IA administration visual loss occlusive stroke stroke (nonspecific) encephalopathy * Data are expressed as percent of patients in each treatment group for whom adverse events were reported at any time during their course. This table includes all 505 patients in the total population. IV = intravenous; IA = intra-arterial; 5-FU = 5-fluorouracil; WBC = white blood cell count. t For chemical analyses, any occurrences of toxicity levels 3 to 5 (on a 0 to 5 scale) are tabulated. tumor progression, but differed in the higher frequency of seizures and the failure of corticosteroid therapy to delay the progression of the encephalopathy in most of the patients. ACT scan revealed hypodensity of the white matter of the perfused hemisphere with mass effect including shift and ventricular compression. These changes were sometimes accompanied by gyral and/or ventricular contrast enhancement progressing to irregular calcification, persistent white matter hypodensity, and ventricular dilation. Magnetic resonance (MR) imaging revealed widespread increased T2- weighted signal intensity in the white matter, along with the mass effect noted on CT scans. The neuropathological findings at autopsy in six such patients were described by Rosenblum, et al. 49 Four brains had no evidence of tumor and a fifth had only a microscopic focus of tumor. Only one patient's brain demonstrated progressive tumor growth. The principal pathological finding was disseminated miliary foci of necrosis with mineralizing axonopathy restricted to the ICA distribution and primarily involving the white matter, which was edematous. In three cases the process had progressed to a histologically dissimilar, massive necrotizing leukoencephalopathy indistinguishable from pure radionecrosis. Discussion Efficacy of lntra-arterial BCNU The most important finding in this study is that intraarterial BCNU administered in doses described does not increase the survival period of newly diagnosed J. Neurosurg. / Volume 76/May

7 W. R. Shapiro, et al. patients with malignant glioma over that afforded by intravenous BCNU. The reasons for the poor results with intra-artefial BCNU are not apparent. The major contributors to the reduced survival with intra-arterial BCNU were the patients with anaplastic astrocytoma. Indeed, one especially tragic result was that patients likely to live longer, those with anaplastic astrocytoma, actually had shorter survival periods if treated with intra-arterial BCNU. Patients with glioblastoma multiforme on average had similar survival rates following BCNU treatment by either route. Despite the devastating nature of the encephalopathy, these patients on average did not die any earlier than the group as a whole. The survival periods of the 16 patients with this complication ranged from 5.6 to 67 months or more, with an approximate median of 13.4 months for the 14 of those patients in the Valid Study Group. Hence, the survival period of these patients did not explain the overall difference in survival times between the intra-arterial and the intravenous BCNU-treated patients. Treatment With BCNU With or Without 5-Fluorouracil The second important finding was that the addition of 5-fluorouracil to BCNU treatment did not enhance survival times. The combination of BCNU plus 5- fluorouracil had been found to be the most effective of four regimes tested earlier by the BTCG (unpublished data), and was also reported to be effective in studies by Levin, et al. 39 The sequential 5-fluorouracil added a cell-cycle-specific agent and was designed to improve the efficacy of BCNU alone. Data from Hoshino, et al., 2s indicated that a 3-day period was sufficient to expose most cycling cells to 5-fluorouracil and that the optimal time for administration of 5-fluorouracil was approximately 2 weeks after BCNU. It has been shown that 5-fluorouracil crosses the blood-brain barrier and therefore significant concentrations could be expected to reach the area of tumor growth; 3~ nevertheless, the drug failed to help. Toxicity of Intra-Arterial BCNU The last important finding was the demonstration of substantial toxicity produced by the intra-arterial BCNU, including blindness and fatal encephalopathy. The clinical course of this encephalopathy and associated C-f-demonstrated changes were first reported in five patients by Mahaley, et al., 44 whose investigation included neuropathologieal study of autopsy material from one of those cases. While such toxicity had been described previously, the high incidence and devastating extent we found had not been predicted by the many reported Phase II trials that preceded ours. The rationale for "regional" chemotherapy via intraarterial drug administration has been well presented by Collins, 9 who noted that drugs such as BCNU with high first-pass extraction and high total-body clearance (rapid metabolism) were especially good candidates for intra-arterial administration. Both higher local drug concentration and reduced systemic toxicity would result from the administration of lower doses of such drugs into the arterial supply of the tumor. Malignant glioma, with its good localization, restricted arterial blood supply, and rare metastatic potential, is an excellent example of a tumor system for which intra-arterial chemotherapy would likely be effective. Levin, el al., 4~ had demonstrated that BCNU concentrations in normal animal brain could be increased fourfold by administering the drug through the carotid artery rather than via intravenous infusion. Fenstermacher and coworkers, ~7'1~ using a mathematical model, predicted that intracarotid administration of BCNU would produce a drug concentration as high as 10-fold greater in brain tissue than would intravenous administration, while bone marrow levels of BCNU were predicted to be equivalent to those following intravenous administration. Early Phase II clinical studies appeared to fulfill the promise of this new technique, with many reported responses and low toxicity. 7"24-~-6'37"42'5~ Eye toxicity consisting of retinitis was reported in patients given cumulative doses of BCNU exceeding 400 mg]sq m. z4':5 While encephalopathy was also described in several studies, its incidence appeared to be related to higher BCNU doses. Our study clearly demonstrated that intra-arterial BCNU was toxic even at modest doses. The difference in our incidence of toxicity and that observed in Phase II studies may relate at least in part to the kinds of patients, the number of intra-arterial infusions, and the timing of BCNU infusion and radiotherapy. In the reported Phase II studies of intra-arterial BCNU administration, most of the patients received only two doses of drug. In contrast, in our study, 34% of the patients actually received at least three courses of intraarterial BCNU. Since the encephalopathy occurred after the second course, we observed more of such cases than had been experienced in the previous Phase II studies. The mechanism of the encephalopathy remains to be elucidated. The neuropathological findings have been published. They consist of various combinations of vascular hyalinization, disseminated miliary foei of necrosis, bland coagulative necrosis, and massive necrotizing leukoencephalopathy. 35,44-49 The lesions are similar to those produced by irradiation, and most patients received radiation therapy concurrent with the intra-arterial BCNU. However, Burger, et al., 4 described similar changes in the brains of four patients who had received high-dose intravenous BCNU therapy for systemic cancer without cranial irradiation. Thus, encephalopathy can occur in the absence of radiation therapy, although it appears to be more common when both therapies are used concurrently. It is of interest in this regard that, in a recent series of 28 newly diagnosed patients treated with intra-arterial BCNU just prior to radiation therapy, two developed fatal leukoencephalopathy. 2 It has been proposed that streaming in the arterial 778 J. Neurosurg. / Vo&me 76 /May, 1992

8 Intra-arterial BCNU for malignant glioma tree produces uneven distribution of a poorly mixed drug, resulting in damaging high focal concentrations in the brain. Lutz, et al., 4~ produced an in vitro model demonstrating streaming from such catheters, especially with slow infusions. Blacklock, et al., 3 described studies of the cerebral distribution of t4c-labeled iodoantipyrine used as a lipid-soluble model of BCNU and delivered into the ICA of rhesus monkeys. Focal nonuniformity of drug delivery was demonstrated with as much as a 13-fold difference in drug concentrations found in anatomically contiguous areas. The authors suggested that the current catheters and slow infusions could account for nonuniform high drug concentrations that would produce focal areas of necrosis. However, this would not explain the large, generally contiguous white matter lesions observed in the pathological specimens. Positron emission tomography (PET) studies by Junck, et al, 3~ using intra-arterial ~SO-H20 in humans demonstrated diffuse distribution of the isotope, not the nonuniform focal hot spots of the monkey studies. Saris, et al., > also used JSO-H20 PET studies, and demonstrated intravaseular streaming in patients given the isotope in the supraophthalmic segment of the ICA, but not in patients given isotope in the infraophthalmic segment (the streaming could be eliminated by diastolephased pulsatile administration); the authors thought the larger diameter of the ICA in the neck reduced the likelihood of streaming. Hence, streaming alone is not likely to be the cause of neurotoxicity but could be a contributing factor to the pathogenesis of the encephalopathy. Methods to Reduce Intra-Arterial BCNU Toxicity Two methods have been reported for reducing eye and brain toxicity from intra-arterial BCNU administration: replacing the alcohol diluent, and supraophthalmic artery drug delivery. Layton, et al., 37 reported that it was possible to eliminate the alcohol by dissolving the BCNU in 5% dextrose and water (DsW); vigorous agitation of the container was required. However, they performed recovery studies of BCNU after DsW dilution and filtering, and found a recovery rate of about 84%. In similar studies at Memorial Sloan- Kettering Cancer Center, recovery after dilution was not as good, being nearer to 50% (B Mehta, unpublished data). Even assuming an 84% recovery rate, the actual delivered dosage would have been 126 mg/sq m. Since the toxicity appears to be at least partially dose-related and earlier studies found no toxicity at 100 mg/sq m, 7 the lack of toxicity is probably consistent with the lower delivered dose. It is of note that, in our study, the sharp reduction in the incidence of encephalopathy occurred when the dose of intra-arterial BCNU was reduced to I00 mg/sq m. Greenberg, et al., 23 subsequently described their results in the treatment of 36 patients with intra-arterial BCNU dissolved in reduced concentrations of alcohol. The initial dose was 200 mg/sq m, but the dosage was escalated to as much as 300 mg/sq m in some courses. Unilateral visual loss occurred in nine patients, bul its frequency decreased after the alcohol concentration was reduced. Four of 12 newly diagnosed patients treated prior to radiotherapy and three of 24 recurrent cases treated after radiotherapy developed unilateral low-density abnormalities in the white matter on CT scans, but only two were symptomatic. Thus, dissolving the BCNU in DsW reduces its toxicity by reducing the delivered dose. However, the purpose of the intra-arterial infusion technique was to increase the tissue concentration fourfold over that produced by intravenous infusion. By halving the intra-arterial dose, the maximum tissue concentration would be reduced to a twofold advantage, hardly worth the additional risk both to the brain and to the eye of an intra-arterial infusion. Another method for reducing the ocular toxicity was to deliver the BCNU into the carotid artery using a flow-directed catheter to carry the tip above the takeoff of the ophthalmic artery] 2'~3 Hochberg, el at., 27 treated 79 patients with 192 infraophthalmic and 66 supraophthalmic infusions, generally after irradiation. Eye toxicity was limited to the patients receiving infraophthalmic infusions. Unfortunately, this did not preclude the occurrence of encephalopathy; in Hochberg's series, 20% of the patients showed CT evidence of white matter disease. In another small series of five patients so treated, four had brain toxicity. 2~ Clayman, el al., ~ treated 15 patients using superselective arterial infusions at a dose of 150 mg/sq m, the drug being dissolved in D~W; no examples of encephalopathy were reported based on CT scans. Similarly, Johnson, el al., 29 treated 20 recurrent glioma patients with intracarotid artery BCNU at a dose of 150 mg/sq m {the drug was dissolved in DsW) and reported asymptomatic low-density areas on CT and one or two cases of visual loss. Thus, selective supraophthalmic artery infusion does not prevent encephalopathy. Intra-Arterial Administration of Other Drugs While it is clear that intra-arterial BCNU at doses likely to be advantageous in regional therapy is too toxic for use in brain tumors, other drugs may be less toxic. Among the agents administered intra-arterially in brain tumor chemotherapy, cisplatin has had the most use. Stewart, el al., s6 described Phase I studies with cisplatin. Both retinal and central nervous system (CNS) toxicity occurred, but only at doses exceeding 60 mg/sq m. Mahaley, el ai., 43 treated 41 patients with recurrent gliomas with monthly intracarotid artery cisplatin at 60 mg/sq m. Of 35 evaluable patients, only one developed neurotoxicity. Newton, el al., ~6 reported more toxicity in their treatment of 12 patients at 58 to 100 mg/sq m: severe CNS toxicity was observed, which included seizures, weakness, aphasia, coma, and visual deterioration. However, 11 of their patients had failed previous intra-arterial BCNU treatment, and the intraarterial combination of the two drugs may be quite toxic. Similar toxicity was reported by Kapp and San ford 32 with combined intra-arterial BCNU and cis- J. Neurosurg. / Volume 76/May,

9 W. R. Shapiro, et al. platin. The BTCG examined intra-arterial cisplatin treatment in a Phase II randomized trial, comparing it to intravenous PCNU (1-2(chloroethyl)-3-(2,6-dioxo-3- piperidyl)-l-nitrosourea). 22 Intra-arterial cisplatin, 60 mg/sq m, was given every 4 weeks in comparison to intravenous PCNU, 100 mg/sq m, given every 8 weeks. In that study, 311 patients were randomly assigned to treatment groups; the overall incidence of encephalopathy associated with intra-arterial cisplatin was 3.9% as compared to 9.5% seen for intra-arterial BCNU in this series (Table 2). Several other drugs have been administered intra-arterially in brain-tumor patients with variable results, including (1-4-amino-2-methyl-5- pyrimidinyl)-methyl-3-(2-chloroethyl)-3-nitrosourea (ACNU), 48'58 PCNU, 55 1-(2-chloroethyl)-l-nitroso-3- (2-hydroxyethyl) urea (HECNU), 16,47 and combination BCNU, cisplatin, and 4'-demethylepipodophyllotoxin 9-(4,6-0-2-thenylidene-~-D-glucopyranoside) (VM- 26). 54 Conclusions The local nature of brain tumors should theoretically permit regional chemotherapy. 9"4 Initial enthusiasm for this technique in small uncontrolled studies has waned with the observation of substantial CNS toxicity using BCNU in large controlled trials. Only well-controlled studies can yield definitive data on the usefulness of such techniques. APPENDIX The institutions, National Institutes of Health grant numbers, principal investigators, and contributing investigators in this project were: Memorial Sloan-Kettering Cancer Center (CA 36047), William R. Shapiro, M.D., George Krol, M.D., Russell W. Walker, M.D.; Duke University Medical Center (CA 36005), Peter C. Burger, M.D.; Montefiore Hospital, University of Pittsburgh (CA 36012), Robert G. Selker, M.D., Melvin Deutsch, M.D.; University of Iowa (CA 36013), John C. VanGilder, M.D.; University of Tennessee (CA 36024), James T. Robertson, M.D.; Indiana University (CA 36027), John Mealey, Jr., M.D.; New York University Medical Center (CA 36014), Joseph Ransohoff, M.D., Maxim Koslow, M.D., Emile N. Hiesiger, M.D., In Sup Choi, M.D.; University of North Carolina (CA 36046), M. Stephen Mahaley, Jr,, M.D., Robert Whaley, M.D.; National Cancer Institute, Sylvan B. Green, M.D. Acknowledgments The authors acknowledge the important contributions of Marie Topor, Julie Buckland, Susan Holmes, and Lauren Rich of the Information Management Services, Inc., for the data collection and processing. Special thanks are extended to the many study coordinators whose tireless devotion to our patients made this study possible. References 1. Aronin PA, Mahaley MS Jr, Rudnick SA, et at: Prediction of BCNU pulmonary toxicity in patients with malignant gliomas. An assessment of risk factors. N Engl J Med 303: , Bashir R, Hochberg H, Linggood RM, etal: Pro-irradiation internal carotid artery BCNU in treatment of glioblastoma multiforme. J Nenrosurg 68: , Blacklock JB, Wright DC, Dedrick RL, et al: Drug streaming during intra-arterial chemotherapy. J Neurosurg 64: , Burger PC, Kamenar E, Schold SC, et al: Encephalomyelopathy following high-dose BCNU therapy. Cancer 48: , Burger PC, Vogel FS: Surgical Pathology of the Nervous System and Its Coverings, ed 2. New York: John Wiley & Sons, Calvo FA, Pastor MA, Dy C, et al: Intra-arterial and intravenous chemotherapy for the treatment of malignant glioma. Preliminary results. Am J Clin Oncol 8: , Cascino TL, Byrne TN, Deck MD, et al: Intm-anerial BCNU in the treatment of metastatic brain tumors. J Neurooncol 1: , Clayman DA, Wolpert SM, Heros DO: Superselective arterial BCNU infusion in the treatment of patients with malignant gliomas. AJNR 10: , Collins JM: Pharmacologic rationale for regional drug delivery. J Clin Oneol 2: , Cox DR: Regression models and life tables. J R Stat Soe (B) 343: , Cutler S J, Ederer F: Maximum utilization of the life table method in analyzing survival. J Chronic Dis 8: , DeBrun GM, Davis KR, Hochberg FH: Superselective injection of BCNU through a latex calibrated-leak balloon. AJNR 4: , DeBrun GM, Vinuela FV, Fox A J, et al: Two different calibrated-leak balloons: experimental work and application in humans. AffNR 3: , Dedrick RL: Arterial drug infusion: pharmacokinetic problems and pitfalls. JNCI 80:84-89, Deutsch M, Green SB, Strike TA, et at: Results of a randomized trial comparing BCNU plus radiotherapy, streptozotocin plus radiotherapy, BCNU plus hyperfractionated radiotherapy, and BCNU following misonidazole plus radiotherapy in the postoperative treatment of malignant glioma, lnt J Radial Oncol Biol Phys 16: , Fauchon F, Davila L, Chatellier G, et at: Treatment of malignant gliomas with surgery, intra-arterial infusions of 1-(2-hydroxyethyl)chloroethyl-nitrosourea, and radiation therapy: a Phase II study. Neurosurgery 27: , Fenstermacher JD, Cowles AL: Theoretic limitations of intracarotid infusions in brain tumor chemotherapy. Cancer Treat Rep 61: , Fenstermacher JD, Gazendam J: lntra-arterial infusions of drugs and hyperosmotic solutions as ways of enhancing CNS chemotherapy. Cancer Treat Rep 65 (Suppl 2): 27-37, Feun LG, Wallace S, Yung WK, et al: Phase-I trial of intracarotid BCNU and cisplatin in patients with malignant intracerebral tumors. Cancer Drug Deliv 1: , Foo SH, Choi IS, Berenstein A, et al: Supraophthalmic intracarotid infusion of BCNU for malignant gliema. Neurology 36: , Green SB, Byar DP, Walker MD, et al: Comparisons of carmustine, procarbazine, and high-dose methylprednisolone as additions to surgery and radiotherapy for the treatment of malignant glioma. Cancer Treat Rep 67: , Green SB, Shapiro WR, Burger PC, et at: Randomized comparison of intraarterial (IA) cisplatin and intravenous (IV) PCNU for the treatment of primary brain tumors (BTCG Study 8420A). Proc Am Soc Clin Oneol 8:86, 780 J. Neurosurg. / Volume 76/May, 1992

10 Intra-arterial BCNU for malignant glioma 1989 (Abstract) 23. Greenberg HS, Ensminger WD, Chandler WF, et al: Intraarterial BCNU chemotherapy for treatment of malignant gliomas of the central nervous system. J Neurosurg 61: , Greenberg HS, Ensminger WD, Seeger JF, et al: Intraarterial BCNU chemotherapy for the treatment of malignant gliomas of the central nervous system: a preliminary report. Cancer Treat Rep 65: , Grimson BS, Mahaley MS Jr, Dubey HD, et al: Ophthalmic and central nervous system complications following intracarotid BCNU (carmustine). J Clin Neuro Ophthaltool 1" , Hochberg FH, Heros DO: Regional infusion for brain tumors, in Lokich JJ (ed): Cancer Chemotherapy by Infusion. Chicago: Precept Press, 1987, pp Hochberg FH, Pruitt AA, Beck DO, et al: The rationale and methodology for intra-arterial chemotherapy with BCNU as treatment for glioblastoma. J Neurosurg 63: , Hoshino T, Barker M, Wilson CB, et al: Cell kinetics of human gliomas, d Neurosurg 37:15-26, Johnson DW, Parkinson D, Wolpert SM, et al: Intracarotid chemotherapy with 1,3-bis-(2-chloroethyl)-l-nitrosourea (BCNU) in 5% dextrose in water in the treatment of malignant glioma. Neurosurgery 20: , Junck L, Koeppe RA, Greenberg HS: Mixing in the human carotid artery during carotid drug infusion studies with PET. J Cereb Blood Flow Metab 9: , Kapp JP, Parker JL, Tucker EM: Supraophthalmic carotid infusion for brain chemotherapy. Experience with a new single-lumen catheter and maneuverable tip. J Neurosurg 62: , Kapp JP, Sanford RA: Neurological deficit after carotid infusion of cisplatin and 1,3-bis(2-chloroethyl)-l-nitrosourea (BCNU) for malignant glioma: an analysis of risk factors. Neurosurgery 19: , Kapp JP, Vance R, Parker JL, et at: Limitations of high close intra-arterial 1,3-bis(2-chloroethyl)- l-nitrosourea (BCNU) chemotherapy for malignant gliomas. Neurosurgery 10: , Kapp JP, Vance RB: Supraophthalmic carotid infusion for recurrent glioma: rationale, technique, and preliminary results for cisplatin and BCNU. J Neurooncol 3:5-11, Kleinschmidt-DeMasters BK, Geier JM: Pathology of high-dose intraarterial BCNU. Surg Neurol 31: , Krull IS, Strauss J, Hochberg F, et al: An improved trace analysis of N-nitrosoureas from biological media. J Anal Toxieol 5:42-46, Layton PB, Greenberg HS, Stetson PL, et al: BCNU solubility and toxicity in the treatment of malignant astrocytomas. J Neurosurg 60: , Levin VA, Chadwick M, Little AD: Distribution of 5- fluorouracil-2-~4c and its metabolites in a murine glioma. JNCI 49: , Levin VA, Hoffman WF, Pischer TL, et al: BCNU-5- fluorouracil combination therapy for recurrent malignant brain tumors. Cancer Treat Rep 62: , Levin VA, Kabra PM, Freeman-Dove MA: Pharmacokinetics of intracarotid artery ~4C-BCNU in the squirrel monkey. J Neurosurg 48: , Lutz RJ, Dedrick RL, Boretos JW, et al: Mixing studies during intracarotid artery infusions in an in vitro model. J Nenrosurg 64: , Madajewicz S, West CR, Park HC, etal: Phase II study -- intra-arterial BCNU therapy for metastatic brain tumors. Cancer 47: , Mahaley MS Jr, Hipp SW, Dropcho EJ, et al: lntracarotid cisplatin chemotherapy for recurrent gliomas.,i Neurosurg 70: , Mahaley MS Jr, Whaley RA, Blue M, et al: Central neurotoxicity following intracarotid BCNU chemotherapy for malignant gliomas. J Neurooncol 3: , Mantel N: Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chemother Rep 50: , , Newton HB, Page MA, Junck L, et al: Intra-arterial cisplatin for the treatment of malignant gliomas. J Neurooncol 7:39-45, Poisson M, Chiras J, Fauchon F, et al: Treatment of malignant recurrent glioma by intra-arterial, infra-ophthalmic infusion of HECNU 1-(2-chloroethyl)-l-nitroso- 3-(2-hydroxyethyl) urea. A Phase I1 study. J Neurooncol 8: , Roosen N, Kiwit JC, Lins E, et al: Adjuvant intraarterial chemotherapy with nimustine in the management of World Health Organization grade IV gliomas of the brain. Experience at the Department of Neurosurgery of Df~sseldorf University. Cancer 64: , Rosenblum MK, Delattrc JY, Walker RW, et al: Fatal necrotizing encephalopathy complicating treatment of malignant gliomas with intra-arterial BCNU and irradiation: a pathological study. J Neurooncol 7: , Safdari GH, Mompeon B, Dubois JB, et al: Intraarterial 1,3-bis(2-chloroetbyl)-l-nitrosourea chemotherapy for the treatment of malignant gliomas of the brain: a preliminary report. Surg Neurol 24: , Saris SC, Blasberg RG, Carson RE, et al: Intravascular streaming during carotid artery infusions. Demonstration in humans and reduction using diastole-phased pulsatile administration. J Neurosurg 74: , Shapiro WR, Green SB: Reevaluatingthe efficacy of intraarterial BCNU. J Neurosurg 66: , 1987 (Letter) 53. Shapiro WR, Green SB, Burger PC, et al: Randomized trial of three chemotherapy regimens and two radiotherapy regimens in postoperative treatment of malignant glioma. Brain Tumor Cooperative Group Trial J Neurosurg 71:1-9, Stewart DJ, Grahovac Z, Benoit B, etal: Intracarotid chemotherapy with a combination of 1,3-bis(2-chloroethyl)-l-nitrosourea (BCNU), cis-diaminedichloroplatinum (cisplatin) and 4'-0-demethyl-l-0-(4,6-0-2-thenylidenefl-D-glucopyranosyl)epipodophyllotoxin (VM-26) in the treatment of primary and metastatic brain tumors. Neurosurgery 15: , Stewart DJ, Grahovac Z, Russel NA, etal: Phase I study of intracarotid PCNU. J Neurooncol 5: , Stewart DJ, Wallace S, Feun L, et al: A Phase I study of intracarotid artery infusion of cis-diamminedichloroplatinum(ll) in patients with recurrent malignant intracerebral tumors. Cancer Res 42: , West CR, Avellanosa AM, Barua NR, et al: Intraarterial 1, 3-bis(2-chloroethyl)-1-nitrosourea (BCNU) and systemic chemotherapy for malignant gliomas: a follow-up study. Neurosurgery 13: , Yamashita J, Handa H, Tokuriki Y, et al: Intra-arterial ACNU therapy for malignant brain tumors. Experimental studies and preliminary clinical results. J Neurosurg 59: , 1983 Manuscript received June 13, Accepted in final form November 13, Address reprint requests to: William R. Shapiro, M.D., Barrow Neurological Institute, St. Joseph's Hospital and Medical Center, 350 West Thomas Road, Phoenix, Arizona J. Neurosurg. / Volume 76/May,